GFM enhancement, do not remove _ (underscore) in words from html id tags

[chadpowers]

I have another addition to the GFM parser. For headers involving involving underscores, i.e. for
### variable_name Github leaves the underscore in the html id attribute, i.e. Github produces id="variable_name" while kramdown createsid="variablename". Can you add this enhancement to the GFM parser to mimic Github's html id generation? Thanks!

[gettalong]

@chadpowers I was looking into this, do you have a live example for this? I.e. source code and resulting HTML produced by Github?

[s3ththompson]

You could perhaps just replace basic_generate_id with the implementation that GitHub uses, for example:

PUNCTUATION_REGEXP = /[^\p{Word}\- ]/u

def basic_generate_id(str)
  gen_id = str.downcase
  gen_id.gsub!(PUNCTUATION_REGEXP, '')
  gen_id.gsub!(' ', '-')
  gen_id
end

I've run into this issue as well. GitHub turns <h1>32-bit Integer</h1> into <h1 id="32-bit-integer">32-bit Integer</h1>, while Kramdown yields <h1 id="bit-integer">32-bit Integer</h1>.

[gettalong]

@s3ththompson Yeah, the problem is that there is no spec for GFM, so we can only guess and use available live examples.

[gettalong]

@parkr Is Jekyll and/or Github Pags using the kramdown parser or kramdown's GFM parser as default? If it is the latter we should probably be careful regarding changes in the ID generation to avoid problems with existing documents.

[parkr]

@gettalong Hi! It uses the GFM parser. Our defaults are kramdown.input: GFM with kramdown.hard_wrap: false. Thanks for checking! 😸

[gettalong]

@parkr Thanks! What do you think: Should we make a hard change now, i.e. adopt Github's ID generation as pointed out by @s3ththompson in #267 (comment) or just change the two mentioned cases (numbers at start and underscores)?

Regarding the needed changes:

I have added some more test cases to a wiki page to see live examples - see https://github.com/gettalong/kramdown/wiki/GFM-test-case.

We will need to set the ID in the GFM parser itself since the normal automatic ID generation is done in the HTML converter.

[parkr]

@gettalong Anything to move the GFM parser closer to actual GFM is 👍 from me. /cc @benbalter @jldec if they have other thoughts.

[gettalong]

@parkr Okay, then I will update the GFM parser so that the ID generation logic from #267 (comment) is used.

[gettalong]

Since we have to do this in the parser (otherwise we would need to create a completely separate HTML converter just for GFM), the results will probably be not completely the same, ie. there may be some cases with different results.

[gettalong]

@chadpowers @s3ththompson @parkr I have implemented this, see commit 257ec15 - can you test this out some more? The test cases from https://github.com/gettalong/kramdown/wiki/GFM-test-case work fine.

I intend to release the next version some time next week.

[gettalong]

Will be in the release that goes out today.